Recap

  • Big difference between two robots
  • unbalance for variable “MESH_FIX” and “PORT”

Recurrent Process

  • drop the unbalance variables
  • use distance matrix
ROBOT_NUM

1 Distance Matrix

The purpose is to evaluate the relationship between patients, and compress their information to the 2-dimensional. While we have some outlier in the data, because there is a patient who take 367 days to discharge after the surgery. We then consider delete it.

1.1 PCA

1.2 t-SNE

In fact, t-SNE is a visualization method, it utilizes simulating low-dimensional data points by t-distributed high-dimensional points.

1.3 Summary

Between “robot” and “senhance-robot”, it seems a little difference between them. Currently, the plots could not display which variables are the main effects.

2 Linear Model

2.1 All variables

summary(lm_robot)
## 
## Call:
## lm(formula = Post_LOS ~ ., data = ROBOT_NUMlm)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.1783 -0.4266 -0.2658  0.0179  7.4986 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  2.90688    0.61283   4.743 2.93e-06 ***
## AGE_group    0.08888    0.08034   1.106  0.26926    
## Sex          0.69262    0.22333   3.101  0.00206 ** 
## BMI_group2   0.05704    0.10855   0.525  0.59958    
## APPROACH    -0.35043    0.11693  -2.997  0.00290 ** 
## ASA2         0.26072    0.16560   1.574  0.11619    
## PRI          0.02425    0.12382   0.196  0.84481    
## TECHNIQUE   -0.71195    0.17490  -4.071 5.66e-05 ***
## Laterality  -0.01849    0.14408  -0.128  0.89797    
## HERNIA_TYPE  0.02821    0.10775   0.262  0.79360    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.056 on 398 degrees of freedom
## Multiple R-squared:  0.1217, Adjusted R-squared:  0.1019 
## F-statistic:  6.13 on 9 and 398 DF,  p-value: 4.295e-08

2.2 Lasso

summary(lm_robotLASSO)
## 
## Call:
## lm(formula = Post_LOS ~ AGE_group + Sex + APPROACH + ASA2 + TECHNIQUE, 
##     data = ROBOT_NUMlm)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.1335 -0.4014 -0.3058  0.0384  7.5352 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  3.08273    0.50377   6.119 2.23e-09 ***
## AGE_group    0.08594    0.07938   1.083  0.27962    
## Sex          0.68066    0.21717   3.134  0.00185 ** 
## APPROACH    -0.35383    0.11512  -3.074  0.00226 ** 
## ASA2         0.25827    0.16367   1.578  0.11536    
## TECHNIQUE   -0.71915    0.16978  -4.236 2.82e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.051 on 402 degrees of freedom
## Multiple R-squared:  0.1208, Adjusted R-squared:  0.1099 
## F-statistic: 11.05 on 5 and 402 DF,  p-value: 5.536e-10

2.3 AIC

summary(lm_robotAIC)
## 
## Call:
## lm(formula = Post_LOS ~ Sex + APPROACH + ASA2 + TECHNIQUE, data = ROBOT_NUMlm)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.0564 -0.3986 -0.3570 -0.0564  7.5572 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   3.2671     0.4742   6.890 2.16e-11 ***
## Sex           0.6525     0.2157   3.026  0.00264 ** 
## APPROACH     -0.3422     0.1146  -2.985  0.00301 ** 
## ASA2          0.3032     0.1584   1.915  0.05624 .  
## TECHNIQUE    -0.7410     0.1686  -4.395 1.42e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.051 on 403 degrees of freedom
## Multiple R-squared:  0.1182, Adjusted R-squared:  0.1095 
## F-statistic: 13.51 on 4 and 403 DF,  p-value: 2.417e-10

2.4 Post_LOS ~ APPROACH

The model illustrates that patients with “senhance-robot” take 0.5 days less than “robot” patients. While it is just a estimation, there is a little bit difference between them.

summary(lm_robot1)
## 
## Call:
## lm(formula = Post_LOS ~ APPROACH, data = ROBOT_NUMlm)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -6.1318 -0.6022 -0.1318  0.3978  8.3978 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   3.0725     0.1625   18.90  < 2e-16 ***
## APPROACH     -0.4704     0.1164   -4.04 6.41e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.094 on 406 degrees of freedom
## Multiple R-squared:  0.03864,    Adjusted R-squared:  0.03627 
## F-statistic: 16.32 on 1 and 406 DF,  p-value: 6.406e-05

3 Conclusions(Now)

3.1 Conclusions

  • The patients with “Senhance-robot” are discharge slightly earlier than others.
  • PCA and t-SNE could not display the difference between two robots.

3.2 Discussion and What next?

  • Split the variables with different information.
  • Tuned the weight distance matrix of different information to display which variable is the most effective.